Willems, Oostenveld, & Hagoort (2008) EEG tends to be oscillatory Composed of several different frequency bands Fourier Decomposition Theta (4-6 Hz), Alpha (8-12 Hz), Gamma (40-50 Hz), … Activity in different frequency bands changes in characteristic ways in response to stimuli/tasks Increase/decrease in amplitude/power Increase/decrease in synchrony across brain areas E.g., Theta becomes more synchronized across regions during memory retrieval
Stimuli Spoken Dutch sentences - Picture presented with a Critical Word - Never sentence-final - CW &/or picture Congruent or Incongruent w/ sentence - 41 trials/cond in 4 conds Task = Listen in order to answer questions later Previous work found N400 effects for both words & pix - Scalp distribution often more frontal for pix Critical Word
More on Stimuli & Procedure Pix appeared at onset of auditory CW and stayed on screen through end of sentence CWs = 1-syll pictureable nouns beginning w/ stop Consonant Incongruent CWs always started w/ different stop C Low constraint context sentences Congruent CW cloze probability = 16% Incongruent CWs cloze = 0% EEG measured from 27 scalp sites Referred to left mastoid, later re-referenced to both mastoids 16 participants
ERP Results Scalp distribution of N400 effect more frontal than for language-only stimuli
Significance Tests of ERP Results Lang Mismatch & Pict Mismatch both evoke N400 of same size & scalp distribution - Both larger than correct - And not diff from each other Double Mismatch also evokes Larger N400 than Correct - But they never say whether it’s smaller than Lang & Pict!
Percent Power Change (from pre-CW baseline) Over Time in EEG Frequency Bands { { theta alpha { gamma
Diffs Between Conditions in Percent Power Change & Their Scalp Distributions { { theta alpha { gamma
Alpha effects about language Gamma effects about picture
Frequency Components of ERP Difference Waves
Some Interpretation Increases in theta power in late window in all mismatch conditions Oscillatory counterpart of N400? Decrease in alpha power in earlier window in both Lang & Pict Mismatch, but stronger in Lang than in Pict Mismatch Maybe early detection that spoken word doesn’t match “predicted” word? Even though cloze is only 16% for congruent words? Decrease in gamma power in earlier time window for Pict Mismatch compared to both Correct & Lang Mismatch Early detection of mismatch in visual info from pict w/ sentence context?